TF-AT: Using the Fallback Runtime to Avoid Slow Training in Colab

NOTE (02/08/2024): The fallback runtime is now unavailable. You can only access it for a few weeks after Colab updates its runtime. Please create a new topic if you’re having training issues with the labs. Thank you.

Google Colab updates its runtime periodically to use newer versions of packages like Tensorflow. This can sometimes affect the use of GPUs for notebooks that require older versions. You will notice that when you reach the training part of the exercise. The lack of GPU will result in waiting several minutes per epoch. To get around it, please use the fallback runtime before running any of the code in your notebook. Here are the steps:

  1. Make sure that you are already connected to a GPU. You can click on that on the upper right of the UI:

  1. After connecting, go to Tools > Command Palette:

  1. Search for “fallback” and select Use fallback runtime version:

  1. Wait for the runtime to reconnect and start solving the assignment (or re-run all your code if you already started it beforehand).

In most cases, you won’t need to use this fallback. But in case you do, please report it by creating a new topic in the Forum so the course mentors and staff will be aware of it. We will try to upgrade the graders to work with more recent versions. This process can take some time so we advise that you use the fallback in the meantime. Thank you!